HelloSun
/

fint-epqv2

Mixture of Experts

expert-offloading

Model card Files Files and versions

fint-epqv2 / src

734 kB

Ctrl+K

Ctrl+K

1 contributor

History: 6 commits

HelloSun's picture

Add auto arena mode (80% of available memory, capped at all experts); patch 369ef971

2873e19 verified 17 days ago

CMakeLists.txt
2.26 kB
Add Flint expert paging port and memory benchmark 17 days ago
llama-context.cpp
170 kB
Add Flint expert paging port and memory benchmark 17 days ago
llama-graph.cpp
159 kB
Add Flint expert paging port and memory benchmark 17 days ago
llama-model.cpp
188 kB
Add Flint expert paging port and memory benchmark 17 days ago
llama.cpp
23.1 kB
Add Flint expert paging port and memory benchmark 17 days ago
sdq_buft.cpp
9 kB
Add Flint expert paging port and memory benchmark 17 days ago
sdq_moe.cpp
55.3 kB
Upload v2 memory tiers (default and forced row-parallel) with path counters 17 days ago
sdq_pager.cpp
105 kB
Add auto arena mode (80% of available memory, capped at all experts); patch 369ef971 17 days ago
sdq_pager.h
23.6 kB
Upload v2 memory tiers (default and forced row-parallel) with path counters 17 days ago